{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:LBR2YVVNQSJX7OPVVKNSI5XVH7","short_pith_number":"pith:LBR2YVVN","schema_version":"1.0","canonical_sha256":"5863ac56ad84937fb9f5aa9b2476f53feb1aad507630a77b584c7b4b7fed07e0","source":{"kind":"arxiv","id":"2305.06324","version":2},"attestation_state":"computed","paper":{"title":"Alternating Gradient Descent and Mixture-of-Experts for Integrated Multimodal Perception","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.MM","eess.IV"],"primary_cat":"cs.CV","authors_text":"Dan Kondratyuk, Hartwig Adam, Hassan Akbari, Huisheng Wang, Rachel Hornung, Yin Cui","submitted_at":"2023-05-10T17:22:06Z","abstract_excerpt":"We present Integrated Multimodal Perception (IMP), a simple and scalable multimodal multi-task training and modeling approach. IMP integrates multimodal inputs including image, video, text, and audio into a single Transformer encoder with minimal modality-specific components. IMP makes use of a novel design that combines Alternating Gradient Descent (AGD) and Mixture-of-Experts (MoE) for efficient model and task scaling. We conduct extensive empirical studies and reveal the following key insights: 1) Performing gradient descent updates by alternating on diverse modalities, loss functions, and "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2305.06324","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2023-05-10T17:22:06Z","cross_cats_sorted":["cs.AI","cs.LG","cs.MM","eess.IV"],"title_canon_sha256":"ba2efb89e8a905cb72868e2b2133a7b9c9dfa02a54d1b42816f09b5409b6d8c8","abstract_canon_sha256":"8814608ed3430a5b43f24eaea5fe6d9948d6889b9a61f21a757615a644cf7056"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:22:17.744565Z","signature_b64":"VJiwoNZGRETNdJ9FwegzRwQc87TlTFopNtRLTePOHQTd9bCkLqfkavouy0WhaCYG/zfd5Lb3Lfz41NlSmwLlDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5863ac56ad84937fb9f5aa9b2476f53feb1aad507630a77b584c7b4b7fed07e0","last_reissued_at":"2026-07-05T07:22:17.744081Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:22:17.744081Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Alternating Gradient Descent and Mixture-of-Experts for Integrated Multimodal Perception","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.MM","eess.IV"],"primary_cat":"cs.CV","authors_text":"Dan Kondratyuk, Hartwig Adam, Hassan Akbari, Huisheng Wang, Rachel Hornung, Yin Cui","submitted_at":"2023-05-10T17:22:06Z","abstract_excerpt":"We present Integrated Multimodal Perception (IMP), a simple and scalable multimodal multi-task training and modeling approach. IMP integrates multimodal inputs including image, video, text, and audio into a single Transformer encoder with minimal modality-specific components. IMP makes use of a novel design that combines Alternating Gradient Descent (AGD) and Mixture-of-Experts (MoE) for efficient model and task scaling. We conduct extensive empirical studies and reveal the following key insights: 1) Performing gradient descent updates by alternating on diverse modalities, loss functions, and "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.06324","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2305.06324/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2305.06324","created_at":"2026-07-05T07:22:17.744141+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.06324v2","created_at":"2026-07-05T07:22:17.744141+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.06324","created_at":"2026-07-05T07:22:17.744141+00:00"},{"alias_kind":"pith_short_12","alias_value":"LBR2YVVNQSJX","created_at":"2026-07-05T07:22:17.744141+00:00"},{"alias_kind":"pith_short_16","alias_value":"LBR2YVVNQSJX7OPV","created_at":"2026-07-05T07:22:17.744141+00:00"},{"alias_kind":"pith_short_8","alias_value":"LBR2YVVN","created_at":"2026-07-05T07:22:17.744141+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2312.14125","citing_title":"VideoPoet: A Large Language Model for Zero-Shot Video Generation","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08209","citing_title":"Learngene Search Across Multiple Datasets for Building Variable-Sized Models","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7","json":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7.json","graph_json":"https://pith.science/api/pith-number/LBR2YVVNQSJX7OPVVKNSI5XVH7/graph.json","events_json":"https://pith.science/api/pith-number/LBR2YVVNQSJX7OPVVKNSI5XVH7/events.json","paper":"https://pith.science/paper/LBR2YVVN"},"agent_actions":{"view_html":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7","download_json":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7.json","view_paper":"https://pith.science/paper/LBR2YVVN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.06324&json=true","fetch_graph":"https://pith.science/api/pith-number/LBR2YVVNQSJX7OPVVKNSI5XVH7/graph.json","fetch_events":"https://pith.science/api/pith-number/LBR2YVVNQSJX7OPVVKNSI5XVH7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7/action/storage_attestation","attest_author":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7/action/author_attestation","sign_citation":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7/action/citation_signature","submit_replication":"https://pith.science/pith/LBR2YVVNQSJX7OPVVKNSI5XVH7/action/replication_record"}},"created_at":"2026-07-05T07:22:17.744141+00:00","updated_at":"2026-07-05T07:22:17.744141+00:00"}